Compression

RocksDB.jl's bindings are built with all five compression backends RocksDB supports. Pick one with the compression keyword, accepted anywhere Options is built (opendb, RocksDBDict, ColumnFamilies, ...):

Toy example: round-tripping through every backend

using RocksDB

for c in (:none, :snappy, :zlib, :bzip2, :lz4, :lz4hc, :zstd)
    opendb(mktempdir() * "/db_$c"; compression = c) do db
        put!(db, "key", "some repeated value " ^ 50)
        @assert String(get(db, "key")) == "some repeated value " ^ 50
        println(c, ": round-trip OK")
    end
end

Compression is applied per SST file block (not per key), so it matters most for larger values and larger databases – a toy example like this mainly confirms that all five backends are actually linked in and functional, which is exactly what test/test_midlevel.jl's own "all compression backends round-trip" test checks.

Why :zstd by default

Zstandard is RocksDB upstream's own recommended default for new applications: close to zlib/bzip2 compression ratios at speeds much closer to Snappy/LZ4. Unless you have a specific reason to change it (e.g. matching an existing database's setting, or :lz4 if write throughput matters more than storage size), the default is the right choice.

Tier 3 equivalent

Every Tier 3 constructor (RocksDBDict, ColumnFamilies) forwards compression (and every other Options keyword) the same way:

using RocksDB

d = RocksDBDict{String,String}(mktempdir() * "/dict_lz4"; compression = :lz4)
d["k"] = "v"
close(d)